the effects of media slant on firm behavior: evidence from ... · affiliated politicians by, for...
TRANSCRIPT
The Effects of Media Slant on Firm Behavior: Evidence from Fox News
Channel Availability and the 2000 Election
Vishal P. Baloria
Boston College
Jonas Heese
Harvard University
November 2016
Abstract: We examine whether the availability of Fox News Channel (FNC) in particular
districts leads firms with ties to the Democratic Party to temporarily suppress negative financial
information during the 2000 election. Using stock return skewness as our measure of negative
financial information, we find that Democratic firms located in districts with FNC availability
experience a reduction (an increase) in negative stock return skewness prior to (after) the
election. We also explore the mechanisms through which Democratic firms suppress negative
information and find that Democratic firms located in districts with FNC availability delay
workforce as well as investment reductions. Collectively, our results, robust to a number of
alternative explanations and possible endogeneity concerns, suggest that media slant can affect
firm behavior in predictable ways.
Keywords: media slant, negative financial information, political connections
Data Availability: All data are available from public sources identified in the paper.
We thank Beth Blankespoor, Nerissa Brown, Mary Ellen Carter, Jeff Cohen, Jenelle Conaway, Emmanuel
De George, David Godsell, Rani Hoitash, Amy Hutton, Thomas Keusch, Krish Menon, Karthik
Ramanna, Sugata Roychowdhury, Ewa Sletten, Eugene Soltes, Ahmed Tahoun, Irem Tuna, Steve Utke,
Florin Vasvari and participants at the Boston Accounting Research Colloquium at Boston University,
London Business School, and the 2016 Accounting Conference at Temple University for helpful
comments. We also thank Rohit Singla, Tristan Norcutt, and Xiaonan Qin for excellent research
assistance. We are grateful to Joshua Clinton and Ted Enamorado for providing data on Fox News
Channel availability.
1
1. Introduction
The media plays an important role in financial markets by disseminating and processing
information about firms. At the same time, media outlets have diverse incentives which can
affect the way they accumulate, cover, and frame news (Miller and Skinner, 2015). One of the
most pervasive institutional features of the contemporary media environment is slant, defined by
Mullainathan and Shleifer (2005) as “the process of selecting details that are favorable or
unfavorable to the subject being described.” Coverage slanted on the basis of political views
(henceforth media slant) can be observed in many different forms of media in the U.S., including
print media (e.g., New York Times, Wall Street Journal), talk radio (e.g., The Ed Shultz Show,
The Rush Limbaugh Show), and television news coverage (e.g., MSNBC, Fox News Channel).
In this study, we examine whether media slant has an effect on firm behavior. We focus
on the Fox News Channel (FNC) because the manner in which it spread across the U.S. allows
us to make predictions on the types of firms that are likely to alter their behavior, the types of
behavior these firms are likely to alter, as well as the time period in which these firms are likely
to alter their behavior.
The national television news media landscape changed significantly in October 1996 with
the introduction of the Fox News Channel, as its owner Rupert Murdoch promised news
coverage slanted to the right of existing television news sources.1 In the years that followed,
FNC spread quickly across the U.S., although cable companies in neighboring towns adopted
FNC in different years, creating idiosyncratic differences in access to FNC. We examine whether
the availability of FNC in particular districts affects the behavior of firms located in those areas.
1 The major television news networks at that time included broadcast (ABC, CBS, and NBC) and cable (CNN)
networks. These four outlets were available to almost all U.S. households. Based on think tank citations data from
1997-2003, Groseclose and Milyo (2005) compute an index of political orientation and find that FNC is significantly
to the right of all other major news media outlets. MSNBC launched in 1996 but had minimal viewership until 2008
when it adopted a distinctively liberal perspective (Groseclose and Milyo, 2005).
2
We focus on firms with ties to the Democratic Party and examine their release of negative
financial information around the 2000 election.
During important political events, politically connected firms cater to the needs of their
affiliated politicians by, for instance, engaging in accrual manipulation (Baloria and Klassen,
2016; Ramanna and Roychowdhury, 2010) or, more generally, suppressing negative financial
information (Piotroski et al., 2015). Politically connected firms engage in such behavior to avoid
negative public scrutiny that might impede their candidates’ political prospects, eliminating the
possibility that these firms will receive economic benefits from their political allies in the future.
Firms are thus willing to take costly actions to preserve valuable relationships with politicians.
During the 2000 election, Fox News openly acknowledged its pro-Bush, anti-Gore
reporting. In fact, after Al Gore announced that Joseph Lieberman would be his running mate,
Mr. Lieberman gave interviews at all of the main television news outlets except for FNC. A
campaign spokesman told FNC that it had been left out because it had “a certain editorial
predisposition” (Rutenberg, 2000).2 FNC also covered Al Gore’s alleged campaign finance
abuses and named firms donating to the Gore campaign. On “Fox News Sunday,” on July 2,
2000, Tony Snow questioned Gore Spokesman Mark Fabiani about Occidental Petroleum’s long-
term support of Gore, criticized one of the firm’s controversial drilling projects in Columbia, and
accused Gore’s policy views to be affected by his financial ties to the company. While other
media outlets also covered this story, Erik Sorenson, MSNBC’s vice president and general
2 As just one example of how FNC openly acknowledged its pro-Bush, anti-Gore reporting, consider the case of
Tony Snow, a former host of Fox News Sunday. During the 2000 campaign, far-right website Free Republic
(www.freerepublic.com) criticized Tony Snow for being critical of the Bush campaign. Tony Snow published a
detailed memo, highlighting the pro-Bush stance at FNC. In particular, he described the strategy he used to
interview a pair of guests about the presidential campaign–John Podesta, an aide to the Democratic Party, and Tom
Ridge, a Republican governor. “We opened with a tough interview of John Podesta, taking Clinton to task for a
series of things and asking some tough questions about Gore’s energy and health-care policies. Tom Ridge came
next. We tried to get him to fire away at Clinton/Gore corruption. He wouldn’t do it. We tried to get him to urge a
more openly conservative campaign by Bush. He wouldn’t do it. If you have complaints about such matters, I
suggest you write the Bush campaign, not Fox News Channel.”
3
manager, said Fox would give those stories more time. Mr. Sorenson (2000) said: “Our approach
is to give it 30 seconds, or a minute. They’re going to talk about it for four hours.” In fact, prior
studies find that FNC’s slanted coverage decreased the vote share for the Democratic Party in
districts with FNC availability in the 2000 election (DellaVigna and Kaplan, 2007; Hopkins and
Ladd, 2014) and made politicians whose constituents had access to FNC less supportive of
incumbent President Clinton (Clinton and Enamorado, 2014).
Following the above arguments, we propose that—because of FNC’s right-leaning
coverage—firms with ties to the Democratic Party and located in areas with FNC availability
have incentives to temporarily suppress negative financial information during the 2000 election.
This expectation is consistent with the right-leaning FNC increasing the costs of bad news
revelation for left-leaning firms, thereby creating an incentive for these firms to take actions to
avoid negative public scrutiny. Following research in accounting, finance, and management (e.g.,
Ayers et al., 2011; Chhaochharia et al., 2012; Hutton et al., 2015), our maintained assumption is
that firms closely monitor their local information environment and alter their behavior
accordingly.3 While FNC is not the only source of information, its clearly stated ideological
position likely made it of particular interest to firms with ties to the Democratic Party.4
Our primary analysis tests this proposition by examining the stock price behavior of firms
with ties to the Democratic Party (Democratic firms) around the 2000 election. We employ a
stock price crash risk methodology to capture the suppression and subsequent release of negative
3 Anecdotal evidence suggests that corporate managers actively consume television news coverage on issues of
interest to their particular firms (e.g., Auletta, 2001). 4 From 1996 to 1999, the Clinton/Gore administration faced considerable scrutiny for their ties to Loral Corporation.
Reports allege the duo accepted significant campaign contributions from Loral executives in exchange for various
policy favors, including facilitating sales of its satellites to China. This became part of the bigger “ChinaGate”
controversy where the administration was accused of leaking secrets that facilitated the growth of the Chinese
nuclear program. In 1999, years after the story initially broke, FNC journalist Carl Cameron declared “ChinaGate is
not forgotten on FNC.” Rep. Curt Weldon (R-PA) noted, “Carl Cameron of Fox News is one of the few dedicated
network reporters who continues to pursue the China story. Like most revelations involving the Clinton-Gore
Administration, the issue was ignored by much of the mainstream press” (Bennet, 1998).
4
financial information in the pre- and post-election year (Piotroski et al., 2015). Within a sample
of Democratic firms, we compare the stock return skewness for firms with FNC availability to
Democratic firms without FNC availability during the time period 1998 to 2003. Firms with
greater negative return skewness are more likely to have large negative stock price movements,
which we interpret as indicative of the release of negative financial information.
We identify firms with ties to the Democratic Party using firms’ soft money
contributions, which are particularly well-suited to identify Democratic firms for our purpose.
Unlike political action committee (PAC) contributions, firms tend to concentrate their soft
money contributions more towards one party rather than (equally) distributing funds to both
parties. Soft money donations are also larger in magnitude as there are no donation limits, and
are donated to the party rather than candidate (Apollonio and Raja, 2004; Jayachandran, 2006).
As in Clinton and Enamorado (2014), we use data collected from the Television and
Cable FactBook on the number of subscribers per Cable Company in each congressional district
with access to FNC. The dataset encompasses FNC availability in 14,748 towns in 35 states for
the years 1998, 2000, and 2003 and reveals that FNC was accessible to 17% of the U.S.
population by June 2000.5 We assign FNC availability to sample firms based on the location of
their corporate headquarters. Prior studies find that the gradual introduction of FNC across the
U.S. was unrelated to the political ideology of voters or incumbents, but was instead primarily
driven by the ease with which Fox was able to negotiate an agreement with local cable
companies (Clinton and Enamorado, 2014; DellaVigna and Kaplan, 2007; Hopkins and Ladd,
2014). Therefore, it is unlikely that our treatment variable (FNC availability) is affected by our
outcome variable (stock return skewness).
5 The 15 omitted states are absent for reasons unrelated to the spread of FNC (Clinton and Enamorado, 2014).
5
We find that Democratic firms located in districts with FNC access release less (more)
negative financial information in the pre-election (post-election) year. This intertemporal pattern
is consistent with the initial suppression and subsequent release of bad news in response to
heightened costs of bad news revelation during the 2000 election. We also examine specific
mechanisms these firms use to suppress negative financial information. We focus on real actions
that are likely to be covered by the media: major employment and investment reductions. Prior
research shows that because current employment and investment levels significantly influence
voters, politically connected firms cater to the needs of their politicians by avoiding layoffs and
plant closures prior to an election (e.g., Bertrand et al., 2007). We focus on large employment
and investment reductions because the market reaction is significantly more negative for material
reductions (McConnell and Muscarella, 1985; Nixon et al., 2004; Worrell et al., 1991), and thus
these reductions are likely to be captured in our stock price skewness measure. Our evidence
suggests that Democratic firms with FNC access are less likely to have large workforce and
investment reductions prior to the election.
Our research design helps reduce internal validity threats as we examine firm behavior
over time, benchmark this behavior to a control group, and exploit intertemporal (i.e., pre-/post-
election) as well as cross-sectional (i.e., FNC availability) variation. The intertemporal variation
controls for time-invariant firm- and district-specific characteristics, whereas the cross-sectional
variation controls for time-varying effects such as the Dotcom bubble, the September 11th
terrorist attacks, or Enron. Despite this, we still conduct a battery of additional tests.
First, we use propensity score matched samples to demonstrate that our results are not
driven by differences in two observable firm characteristics that could plausibly affect our
inferences: strength of ties to the Democratic Party (as measured by the amount of soft money
6
contributions) or firm visibility (as measured by analyst following, Fortune 500 membership, and
the incidence of restatements).6
Second, we use a falsification test (where we try to replicate our
results around the 1996 election) and a firm fixed effects model to demonstrate that our results
are not driven by unobservable characteristics of firms or geographical areas. Third, we rule out
the possibility that FNC access mainly captures information availability rather than media slant
by controlling for information availability (as measured by access to broadband internet) and
rerunning our tests on a sample of Republican and non-politically connected firms.7 Finally, we
run several alternative specifications to reduce measurement error in our empirical proxies for
media slant and ties to the Democratic Party.
Our study contributes to the literature in several ways. One contribution is to the
reemerging political costs literature (Watts and Zimmerman, 1978), which uses various political
events to test the political cost hypothesis (Kido et al., 2012; Piotroski et al., 2015; Ramanna and
Roychowdhury, 2010). Our study adds to this literature by exploiting a pervasive feature of the
U.S. political landscape, media slant, to address Miller and Skinner’s (2015) explicit call for
research on the role played by the media in the political process.
Our study also contributes to research that examines the effects of media coverage in
financial markets. Prior research suggests that the media can play an information, monitoring, or
catering role, and has largely focused on the business press (Miller and Skinner, 2015). Across
all roles, there is considerable evidence that the choice of firms to cover and the resulting news
6 A more direct measure of firm visibility could be the number of newspaper articles per firm. However, to the
extent that FNC availability drives newspapers to report about firms, such measure is biased. Instead, we focus on
firm characteristics that influence firm visibility but are less affected by FNC. 7 Relative to our sample of Democratic firms, these firms should also be affected by the information availability
channel but unaffected by the media slant channel. For these firms, we find that FNC availability does not
significantly change firm behavior in the period prior to or subsequent to the 2000 election, inconsistent with the
information availability explanation. These tests imply that unobservable and omitted firm or geographic
characteristics would have to affect only Democratic firms, do so differently over time, and be correlated with, but
not driven by, FNC availability.
7
reported on these firms are not random decisions but affected by the incentives of the media
(e.g., Mullainathan and Shleifer, 2005). We use the diffusion of FNC and its overlap with the
2000 election to document that media slant can have a measureable effect on firm behavior. The
setting we examine is specific to a particular time period and media outlet, thereby requiring to
exercise caution in generalizing the findings. Nonetheless, the setting offers a powerful setting to
provide evidence on the effect of media slant on firm behavior.
Finally, our study extends the examination of FNC consequences to financial markets.
Prior research in economics and political science finds that FNC caused a shift towards the right
among voters (DellaVigna and Kaplan, 2007; Hopkins and Ladd, 2014) as well as politicians
(Clinton and Enamorado, 2014). Our study sheds light on how developments in media markets
can also influence the behavior of economic agents in financial markets.
The remainder of this paper is organized as follows. Section 2 reviews prior literature,
provides institutional background on the setting, and presents our hypothesis. Section 3 describes
our sample and research design. Section 4 discusses our results. Section 5 concludes the paper.
2. Literature Review, Background, and Hypothesis
2.1. The Role of the Media in Financial Markets
Understanding the incentives of the media in its coverage of corporations is a developing
area of research. One view of the media is that it plays an important monitoring and information
role in financial markets. As described in Miller and Skinner (2015), the extant research in
support of this view has largely focused on two questions: (1) does the media play a monitoring
role in financial markets?; (2) does the media play an information role and affect stock price
formation through provision of value-relevant information? There is evidence to support the
media’s monitoring role in bringing corporate fraud to light (Miller, 2006), curbing excess
executive compensation (Core et al., 2008; Kuhnen and Niessen, 2012), and insider trading (Dai
8
et al., 2015; Rogers et al., 2016). There is also evidence to support the information role as media
coverage affects stock price formation by reducing information asymmetry (Bushee et al., 2010),
cash flow mispricing (Drake et al., 2014), and thereby stimulating trading activity around
earnings announcements (Engelberg and Parsons, 2011).
At the same time, prior research has argued that the media also caters to consumers
(Jensen, 1979). Evidence of media biases relating to consumers’ political preferences provide
support for this view (Gentzkow and Shapiro, 2010; Mullainathan and Shleifer, 2005). The
media has also been shown to face conflicts of interests in its coverage of firms with which it has
advertising relationships (Gurun and Butler, 2012; Reuter and Zitzewitz, 2005).
As described in more detail in the literature reviews by Cohen et al. (2015) and Miller
and Skinner (2015), this body of research suggests that the media faces a number of incentives
that can shape the manner in which it accumulates and reports news about firms.
2.2. Political Events and the Information Environment
Watts and Zimmerman’s (1978) political costs hypothesis posits that the desire to
preserve positive—or to avoid negative—interactions with the government provides managers
with incentives to manage accounting reports around political events. Ramanna and
Roychowdhury (2010) extend this hypothesis to an electoral setting by showing that politically
connected firms identified as being outsourcers manage earnings downwards during the 2004
election, when outsourcing was a major election issue. Kido et al. (2012) and Baloria and
Klassen (2016) provide further evidence consistent with the electoral based political costs
hypothesis by focusing on specific accrual choices made by managers in response to political
incentives. Piotroski et al. (2015) contribute to this literature by broadening the outcome variable
from accrual choices to the overall information environment (as measured by stock price crash
risk). They focus on China where the political events are meetings of the National Congress of
9
the Chinese Communist Party and promotions of high-level provincial politicians. Given the
pervasiveness of political incentives in the Chinese setting, the authors focus on all Chinese
listed firms (rather than the subset of firms that are politically connected).
A key stakeholder in all of these studies is the media. In Ramanna and Roychowdhury
(2010), firms are concerned about negative outsourcing related scrutiny from the media and the
authors consequently use a newspaper-based source as their outsourcing proxy. In Baloria and
Klassen (2016), firms are concerned about effective tax rate related scrutiny from the media and
watchdog groups and the authors demonstrate that firms are less likely to behave
opportunistically when monitoring by the media is strong. Piotroski et al. (2015) complement
their evidence of greater suppression of negative financial information by demonstrating that the
number of newspaper articles about sample firms decreases around political events, suggesting
that in China politicians can censor the media. While this literature acknowledges the important
role of the media in imposing political costs on firms (or shielding firms from these costs), it has
not thoroughly examined the incentives of the media, particularly around important political
events (Miller and Skinner, 2015).
2.3. The Effect of Media Slant and Background on the Fox News Channel
In this paper, we focus on the influence of media slant on firm behavior. Examining such
influence is challenging because media availability is not a random choice, but rather the product
of profit maximization and other utility maximizing objectives. An identification problem arises
because the same unobservable factors that influence the presence of media slant can influence
behavior of economic agents such as managers and their firms. To study the influence of media
slant on firm behavior, the preferable research design would be an experiment in which firms
were randomly assigned ideologically slanted coverage. Such an experiment would allow the
10
researcher to compare treated and non-treated firms’ outcomes and attribute any differences to
the media source. To relate media slant to political costs, the timing of the random assignment
would have to coincide with a political event important enough to generate sufficient incentives
for firms to take actions to avoid political costs.
We maintain that the entry of Fox News Channel in 1996 in local cable markets is close
to such an experiment. Prior to the launch of Fox News Channel, a twenty-four-hour-a-day cable
news channel, the Fox Broadcasting Corporation offered no national news broadcasts
(DellaVigna and Kaplan, 2007). Rupert Murdoch and Roger Alies, Chairman and CEO of Fox
News, launched FNC in October 1996 with the objective of providing a distinctive news
perspective. Based on data from 1997-2003, Groseclose and Milyo (2005) compute an index of
political orientation of news programs and document that FNC is significantly to the right of
other major media outlets such as television (ABC, CBS, CNN, and NBC), online (Drudge
Report), print (New York Times, Newsweek, USA Today, Wall Street Journal, Washington Post,
Washington Times), and radio media outlets (NPR). The availability of ideologically slanted
media alters the political information available to voters with access to that media source. In fact,
studies document that FNC availability decreased the vote share for the Democratic Party in the
2000 election (DellaVigna and Kaplan, 2007; Hopkins and Ladd, 2014).
At the time of the launch, Rupert Murdoch’s stated goal was to make FNC available to as
many people as possible to maximize ratings and revenue (DellaVigna and Kaplan, 2007).
Therefore, the availability of FNC by 2000 in a particular town is primarily based on the ease
with which Fox was able to negotiate an agreement with local cable companies. FNC had to
convince local cable companies to add the channel to its lineup in lieu of another channel. Most
towns have only one local cable company, and the company faces a technological constraint on
11
the number of channels it can offer. FNC offered cable companies a one-time payment of $10
per subscriber to incentivize them to include FNC in their programming. The timing of FNC’s
agreements with local cable companies varied, creating idiosyncratic variation in the diffusion of
FNC.8 We use this random assignment to identify the effect of media slant on firm behavior.
Prior studies provide evidence that the availability of FNC in a particular geographic area
is largely random over this time period. For instance, Clinton and Enamorado (2014) find that
the availability of FNC in a particular district by the year 1998 and 2000 is not related to political
characteristics of the districts’ House representatives or voters. Hopkins and Ladd (2014)
conduct an exhaustive set of tests and note that Fox News’ expansion “was not politically
driven,” but Fox News availability tended to be concentrated in larger U.S. towns with more
cable channels. Within our sample of Democratic firms, we find that the average population in
congressional districts with and without FNC access is very similar (i.e., 614,723 and 631,093,
respectively), suggesting that the district population is unlikely to influence our inferences.
Nonetheless, we address the concern that systematic differences across districts with and without
FNC access might affect our results in Section 4.
In the years that followed its 1996 launch, FNC spread quickly, and by the year 2000,
FNC was available in 22 of the 35 states and accessible to 17% of the U.S. population. Figure 1
shows a map of FNC availability in 2000 across the U.S. Important for our study, the spread of
FNC coincides fairly closely with the 2000 election, increasing the likelihood that firms are
willing to take actions to avoid public scrutiny.
– Please insert Figure 1 about here –
2.4. Hypothesis
8 For example, TCI Cable signed an early agreement with FNC and, by the year 2000, included FNC as part of its
programming in 33% of the towns it served. In contrast, Adelphia Communications signed a late agreement with
FNC and, by the year 2000, included FNC as part of its programming in only 8% of the towns it served.
12
We focus on the effects of media slant on the information environment of U.S. firms
around the 2000 election. Prior research finds that the availability of FNC increased the vote
share for the Republican Party in the 2000 election because the media outlet’s right-leaning
coverage influenced voting outcomes (DellaVigna and Kaplan, 2007; Hopkins and Ladd, 2014).
Clinton and Enamorado (2014) demonstrate that politicians whose constituents had access to
FNC rationally anticipated the ideological shift towards the Republican Party and became less
supportive of incumbent President Clinton. Following this line of research, we examine whether
firms with ties to the Democratic Party take actions to avoid public scrutiny arising from a new
and ideologically slanted information source in their congressional district.
FNC focuses on national, rather than local, issues and thus could still cover the behavior
of firms located in districts in which it is not available. This would confound our ability to detect
differences as all Democratic firms would have incentives to suppress negative financial
information, regardless of FNC availability.9 However, we expect the effect to be strongest for
Democratic firms located in areas with FNC availability. Prior research documents that local
stakeholders have a greater ability to influence firm behavior and as a result, firms pay
particularly close attention to developments in their local information environment.10
For
instance, the preferences of local stakeholders have been shown to influence the type of unethical
behavior firms partake in (Hutton et al., 2015), firms’ governance structures (Chhaochharia et al.,
2012), and use of financial reporting discretion (Ayers et al., 2011). These studies maintain that
because local stakeholders encounter lower costs in communicating with managers, firms are
9 We find no evidence to support this conjecture. Across all of our tests outlined in Section 4, we find no evidence
that Democratic firms without FNC availability suppress and subsequently release negative financial information. 10
In Section 4, we show that our results are robust to restricting our sample to geographically concentrated firms.
13
particularly responsive to the needs of stakeholders in their local environment.11
In the context of
elections, politically connected firms are likely to avoid negative public scrutiny that might
impede their candidates’ political prospects, as firms’ economic benefits from their political
allies depend on their candidate being elected (Cooper et al., 2010).
We also acknowledge that FNC is not the only source of information and other media
outlets could also provide negative coverage of Democratic firms. However, because FNC
explicitly stated its ideological position, delivered it by the most readily consumed news outlet at
the time (i.e., television), and influenced voters (DellaVigna and Kaplan, 2007), we argue it was
likely to be of particular interest to managers.
We propose that—because of Fox News Channel’s right-leaning coverage—the
availability of Fox News Channel in particular districts creates incentives for firms located in
these districts and with ties to the Democratic Party to temporarily suppress negative information
during the 2000 election, and subsequently release such information after the election. This
incentive arises because we expect that the revelation of negative information by firms closely
linked to the Democratic Party would be heavily publicized by FNC, and adversely affect the
future prospects of both the Democratic Party and allied firms. We also expect the temporarily
suppressed negative financial information to be released in the period following the election. Our
hypothesis can be summarized as follows:
Hypothesis: Democratic firms located in districts with FNC availability temporarily
suppress negative information before the 2000 election and subsequently release it
following the election.
While we argue that our setting exhibits a number of features which increase the
likelihood of observing firms responding to the potential public scrutiny created by the
11
Managers of Democratic firms are also more likely to be aware of the potential costs that FNC could impose on
their firms and the Democratic Party if they live and work in an area where FNC is televised.
14
confluence of FNC diffusion and the 2000 election, we note that our hypothesis is not without
tension. First, Democratic firms may not perceive FNC, a new and relatively unproven entrant in
the media market, as a credible threat to the electoral prospects of the Democratic Party. Second,
even if Democratic firms perceive FNC as a credible threat, these firms may not be concerned
about the imposition of political costs as attributing the behavior of an individual firm to the
Democratic Party is a difficult task for any media outlet.12
Finally, deferral of negative financial
information can be costly, thereby requiring that the benefits associated with such behavior are
sufficiently high to offset the costs.13
Therefore, the influence of FNC on firm behavior around
the 2000 election is ultimately an empirical question.
3. Sample, Research Design, and Descriptive Statistics
3.1. Sample
We obtain data on Fox News Channel availability and the number of subscribers per
congressional district from Clinton and Enamorado (2014), who built upon the data collected by
DellaVigna and Kaplan (2007). DellaVigna and Kaplan (2007) use the Television & Cable
Factbook to collect the number of FNC subscribers for 9,256 towns in 28 states, but could not
identify the congressional district for an additional 5,462 towns. Clinton and Enamorado (2014)
use the Congressional District Atlas for the 103rd Congress to identify the congressional district
(or districts) for these additional towns which extends the data to districts in Florida, Delaware,
Indiana, Illinois, Oklahoma, Oregon, and Maryland.14
This leads to a final dataset of FNC
12
Fox did not formally enter into business news until the launch of Fox Business Network (FBN) in 2007. Prior to
FBN, business news was discussed on FNC to the extent that it related to broader political or societal issues.
Bonaparte and Kumar (2013) document a positive association between FNC availability and stock market
participation rates, which they interpret as suggestive of a link between mainstream media and financial markets. 13
Examples of specific costs are higher cost of capital, higher transaction costs, and lower contractual efficiency. 14
For towns in multiple districts they assume that FNC is uniformly distributed. We make the same assumption.
15
availability and subscribers in 14,748 towns in 35 states for the years 1998, 2000, and 2003.
Thus, we restrict our sample to firms headquartered in one of the 35 states with FNC data.15
Our sample is concentrated around the 2000 election and covers the years 1998-2003. We
begin with 1998 because this is the first year for which data on Fox News Channel subscriptions
is available. The sample ends in 2003 to ensure that our sample period does not extend into the
2004 election. We exclude foreign firms cross-listed in the U.S. because foreign firms are legally
not allowed to influence electoral outcomes in the United States and thus cannot make soft
money contributions or create a Political Action Committees (PAC) (Milyo et al., 2000).
After applying data filters, we have 18,649 firm-year observations, representing 4,825
distinct firms as shown in Table 1. Our main sample is comprised of 670 firm-year observations
of firms contributing soft money to the Democratic Party, representing 186 distinct firms.
– Please insert Table 1 about here –
3.2. Research Design
We use the following regression model to examine the relation between FNC availability
and suppression of negative financial information around the 2000 election (where the subscript i
represents the firm and t the year):
Ncskewi,t = FNCi,t + Electioni,t +Post_Electioni,t +FNCi,t x
Electioni,t +FNCi,t x Post_Electioni,t + nn Controlsi,t +i,t
Ncskew is our measure of firm-level variation in the flow of negative information into
stock prices. We use the crash risk methodology developed by Chen et al. (2001) and Jin and
Myers (2006). We follow Piotroski et al. (2015) and employ this methodology in an event period
setting. Specifically, we use the negative skewness statistic which captures the presence of large,
15
In particular, we do not have data on FNC availability for the following states: Arizona, Colorado, Georgia,
Kansas, Kentucky, Louisiana, Mississippi, Montana, North Carolina, Nebraska, New Mexico, Nevada, Texas,
Washington, and West Virginia.
16
negative stock price movements. Firms with greater negative skewness are more likely to have
large negative stock price movements, which we interpret as indicative of the release of material
negative financial information. Ncskewi,t is measured as the third moment of each stock’s weekly
residual returns, divided by the cubed standard deviation of weekly residual returns, times
negative one, in year t. For each firm-year, we assign weekly returns to the 12-month period
ending three months after the firm’s fiscal year-end. This ensures that relevant financial data is
available to market participants. We use one-year intervals because the skewness statistic is
required to be measured over longer periods.16
By multiplying the statistic by negative one, we
can interpret an increase in Ncskewi,t as reflecting an increase in the release of negative financial
information. Similarly, we interpret a decrease in Ncskewi,t as reflecting the suppression of
negative financial information. Ncskewi,t is calculated using the residuals from annual, firm-
specific estimations of the following market model using firm i’s weekly returns in year t:
ri,t = 0 + 1 rm,t-2 +2 rm,t-1 + 3 rm,t +4 rm,t+1 +5 rm,t+2 + i,t
where ri,t is the return of firm i in week tand rm,t is the return on the CRSP value-
weighted market index in week tWe include the lead and lag terms for the market return to
allow for trading frictions and nonsynchronous trading.
FNC is either the logged number of subscribers with FNC access in the congressional
district in which firm i is headquartered in year t or an indicator equal to one if firm i is
headquartered in a district with FNC availability, and zero otherwise. While the first FNC
measure captures the extent of FNC access, the latter captures the existence of FNC access.
Within areas with FNC access, the average number of subscribers is 191,372.
16
Piotroski et al. (2015) examine this issue thoroughly by supplementing their annual specifications with quarterly
specifications and find stronger results in their annual specifications. In addition, our subsequent tests also rely on
Compustat data on the number of employees, which is only available on an annual basis.
17
Election is equal to one for all firms with fiscal year ends between July 31, 1999 and June
30, 2000, i.e., the firm-year before the political event, and zero otherwise. This time window
ensures that firms’ financial statements are likely to be released before Election Day, i.e.,
November 7, 2000. Post_Election is equal to one for all firms with fiscal year ends between July
31, 2001 and June 30, 2002, i.e., the firm-year after the political event window, and zero
otherwise. We define fiscal year 2001 (rather than fiscal year 2000) as the Post_Election variable
because Piotroski et al. (2015) find that the suppression of negative financial information by
firms persists for up to three months after the end of a political event. They argue that because
key policy decisions are typically made during this time period, the potential for scrutiny persists
for several months after the end of the political event. As the outcome of the 2000 presidential
election was not known until December 13, 2000 because of the Florida recount, this effect is of
particular importance in our setting. Figure 2 summarizes our empirical design.
– Please insert Figure 2 about here –
The interactions between FNC with Election and Post_Election capture the incremental
effect of FNC on the incentives to suppress (release) negative financial during the period prior to
(subsequent to) the 2000 election. A decline in negative skewness during our event window, i.e.,
a negative and significant coefficient , is interpreted as a temporary reduction in the flow of
negative information about these firms. A subsequent increase in negative skewness in the period
following the event window, i.e., a positive and significant coefficient , is interpreted as the
release of the previously suppressed negative financial information.
We estimate Model 1 using OLS from 1998-2003 for a sample of 670 firm-year
observations of Democratic firms. We identify firms with ties to the Democratic Party by
collecting data on firms’ soft money contributions to the Democratic Party from the Center for
18
Responsive Politics (CRP). We include the natural logarithm of firms’ soft money contributions
to the Democratic Party in our model, denoted Log_Soft_Money_Dem, to control for differences
in contribution levels.
Following Chen et al. (2001), Jin and Myers (2006), and Piotroski el al. (2015), we also
include a number of firm characteristics to control for other determinants of negative skewness.
As larger firms are less likely to experience dramatic price declines, we control for firm size
(Log_Mark_Cap), measured as the natural logarithm of a firm’s market value of equity at the end
of year t. Growth results in greater susceptibility to price crashes, and therefore we control for
firm growth (Chg_Sales), measured as the sales growth in year t. As the systematic and
idiosyncratic risk characteristics of the firm are negatively associated with negative skewness,
we control for the standard deviation of weekly residual returns (Sigma) and the firm’s beta
(Beta). Stocks with heightened trading volume are more prone to crashes, therefore we control
for contemporaneous share turnover (Turnover), measured as the average weekly share turnover
in year t. As performance may influence the propensity for a stock to crash, we control for
market performance (Return), measured as the annual market-adjusted stock return in year t. In
addition to contemporaneous values of Turnover and Return, we also include lagged measures of
these variables to control for momentum and behavioral factors. We also include state, year, and
two-digit SIC code industry fixed effects to control for idiosyncratic time, state, or industry
factors that can influence crash risk. Following Piotroski et al. (2015), we cluster standard errors
at the state level to capture variation in standard errors across states.
Several features of our research design are worth noting. First, because we examine
differences in firm behavior across time, aspects of the congressional district or firm that do not
change over time cannot drive the observed change in firm behavior. Second, because we
19
benchmark the behavior of Democratic firms with FNC availability (treatment firms) to
Democratic firms without FNC availability (control firms), our research design controls for time-
varying effects. For example, to the extent that our sample period coincides with significant
macroeconomic events (e.g., Dotcom bubble, September 11th
terrorist attacks, Enron, the passage
of SOX, etc.), both sets of firms should be affected by such events. Third, the intertemporal (i.e.,
pre-/post-election) and cross-sectional (i.e., FNC availability) variation in our setting reduces
endogeneity concerns. Finally, measurement error in our empirical proxies for information
suppression, Democratic firms, or FNC availability reduce the power of our empirical tests and
bias against finding evidence consistent with our hypothesis.
3.3. Descriptive Statistics
Table 2 reports descriptive statistics for the variables described above. Panel A presents
statistics for the complete pooled sample and the sample of Democratic firms separately. Panel B
compares means between firms with FNC and without FNC availability for the sample of
Democratic firms, and also provides means for Democratic firms headquartered in the 15 states
for which FNC data is not available. Panel A of Table 2 shows that, in the pooled sample,
approximately 4 percent of firms make soft money contributions to the Democratic Party,
spending on average about $52,706. Within the sample of Democratic firms, 68 percent of the
soft money contributions are made to the Democratic Party. In the pooled sample, approximately
47 percent of firms are headquartered in a district with FNC availability. In the sample of
Democratic firms, approximately 53 percent of firms are located in a district with FNC
20
availability. Democratic firms are on average larger than firms in the pooled sample, which is
consistent with firm size being a significant determinant of political activity.17
Panel B shows that Democratic firms located in districts with FNC availability contribute
significantly more than those located in districts without FNC availability, reinforcing the
importance of controlling for soft money contributions in our tests. Democratic firms located in
districts with FNC availability also have a higher turnover and higher beta. However, there are
no significant differences in any of the other firm characteristics used in our empirical model,
suggesting that, by and large, Democratic firms located in areas with and without FNC
availability are comparable. There are also statistically significant but economically small
differences in the demographics of districts with and without FNC availability. In particular, in
areas with FNC availability, the total population is slightly smaller, the household income is
somewhat lower, and the unemployment rate is 0.4 percentage points higher. This potentially
raises the concern that systematic differences in the underlying economics across districts with
and without FNC availability might affect our inferences. We address these issues in Section 4.
– Please insert Table 2 about here –
Table 3 presents the correlations coefficients for our sample of Democratic firms. Both
measures of FNC are highly correlated with each other (0.91). Soft money contributions are
correlated with firm size (0.35 to 0.44), reinforcing the idea that larger firms contribute more.
– Please insert Table 3 about here –
4. Results
4.1. Main Results
17
While FNC was available to only 17% of the U.S. population by June 2000, about 50% of our sample firms are
headquartered in areas with FNC availability. This is likely explained by the fact that corporate headquarters are
geographically clustered in and around major urban centers, most of which had FNC availability by the year 2000.
21
Table 4 shows the results of estimating equation (1) for our sample of Democratic firms.
We find a negative and significant coefficient on the interaction between Election and FNC, and
a positive and significant coefficient on the interaction between Post_Election and FNC using
both FNC measures. These coefficients are statistically significant at the 5% level or higher (one-
tailed). The coefficient estimates on the two interaction variables in column (1) are 0.037 and
0.030, respectively. The economic magnitudes of these coefficient estimates imply a significant
effect as Chen et al. (2001) find that a change in negative skewness of 0.037 yields an
economically meaningful increase in the price of put options. The coefficient estimates in
column (2) also imply that the observed difference between Democratic firms with and without
FNC availability represents an economically meaningful (i.e., approximately one-third standard
deviation) reduction in negative stock return skewness during the pre-election period. The
subsequent increase in negative stock return skewness in the post-election period is of similar
magnitude, consistent with the release of withheld negative financial information. None of the
coefficients on FNC, Election or Post_Election are significant, suggesting that Democratic firms
do not generally suppress, and subsequently release, negative information in the period
surrounding the 2000 election. These results indicate that Democratic firms located in districts
with FNC access suppress negative information prior to the election and subsequently release
that information after the election, providing support for our hypothesis. The coefficients on the
control variables, where significant, are consistent with prior research. The lack of significance
on the majority of control variables is to be expected given the fact that the two groups of firms
are fairly similar, our sample size is relatively small, and our inclusion of state, industry, and
year fixed effects likely limits variation.
– Please insert Table 4 about here –
22
4.2. Differences in Observable Firm Characteristics
A potential concern with the results in Table 4 is that Democratic firms located in areas
with and without FNC availability differ across several observable firm characteristics. For
instance, Democratic firms located in areas with FNC availability have higher soft money
contributions and may also be more visible. Therefore, these firms might be more concerned
about public scrutiny around the election–independent of FNC. While we control for soft money
contributions in our main test, we conduct additional empirical tests to address concerns that our
inferences are confounded by differences in observable firm characteristics.
Differences in Contributions. First, we create a matched sample of Democratic firms
located in districts with and without FNC availability. In particular, we match to each
Democratic firm with FNC availability a Democratic firms without FNC availability from the
same year, two-digit SIC code industry, size, and soft money contributions using a propensity
score matching method. We match FNC and non-FNC firms within a predefined propensity
score radius (or “caliper”) of 0.15.18
We allow for replacement in the selection of matches to
ensure that we find a meaningful match for each of our FNC firms (Shipman et al., 2016).19
For
the 356 Democratic firm-year observations located in districts with FNC access, we find matches
for 183 observations yielding a total sample size of 366. Table 5, Panel A presents multivariate
results for the matched-sample regressions. The model in Table 5, Panel A is similar to Table 4
except for the fact that firm size, soft money, as well as industry and year fixed effects are
excluded because of the matching on these variables. The results in Table 5, Panel A are similar
18
This radius ensures that we form pairs that do not exhibit significant differences on our chosen covariates. 19
Shipman et al. (2016) argue that matching without replacement may result in lower quality matches and smaller
sample size than matching with replacement, as each control observation may be matched only once, even if it is the
best match for several treatment observations. Thus, replacing observations reduces bias because each treated
observation matches with the most similar control observation. In untabulated tests, we repeat the matching process
and do not allow for replacement. If we match without replacement, we end up with a matched sample of 230 firm-
year observations. Our results are similar using this smaller sample.
23
to those in Table 4 in that they suggest that Democratic firms located in FNC districts suppress
negative information before the election, and subsequently release it after the election.
Differences in Firm Visibility. Another potential concern with the results in Table 4 is
that Democratic firms located in areas with FNC availability might be more visible and thus have
greater incentives to suppress negative information. While we control for firm size, which is
often used as a proxy for firm visibility (Watts and Zimmerman, 1978), in our main test, we use
two different empirical strategies to address this concern. First, we rerun Model 1 and include
three additional control variables for firm visibility. In particular, we include a Fortune_500
indicator, the natural logarithm of the number of analysts issuing annual earnings forecasts for
firms covered by IBES, denoted Log_Analyst, and a Restate indicator, capturing firms’
restatements that might increase firm visibility. As shown in Table 5, Panel B, Column 1 and 2,
our results are robust to including these additional controls for firm visibility. Second, we create
a matched sample of Democratic firms located in districts with and without FNC availability. We
match to each Democratic firm with FNC availability a Democratic firm without FNC
availability from the same year, two-digit SIC code industry, size, Fortune 500 membership,
analyst following, and occurrence of a restatement using a propensity score matching method.
For the 356 Democratic firm-year observations located in districts with FNC availability, we find
matches for 189 observations yielding a total sample size of 378. As shown in Column 3 and 4 of
Table 5, Panel B, we find results consistent with those in Table 4.
– Please insert Table 5 about here –
4.3. Differences in Unobservable Firm Characteristics
Another potential concern with the results presented thus far is that Democratic firms
located in areas with FNC availability may have unobservable characteristics that drive the news
suppression behavior we document. For instance, these firms may be subject to unfavorable
24
economic situations that drive their higher news suppression. We control for this possibility in
the regressions described above through state fixed effects. To further alleviate this concern, and
more broadly the concern that unobservable traits of Democratic firms located in areas with FNC
availability drive news suppression, we conduct two tests.
The 1996 Election. First, we rerun Model 1 for the 1996 election. As FNC was first
introduced in October 1996 but not available for most U.S. households, we can use the 1996
election to examine whether firms with ties to the Democratic Party located in areas with FNC
availability in 2000 behaved similarly during the 1996 election when FNC was not yet available.
We run this test for two sets of firms: 1) the sample of firms that contributed to the Democratic
Party during the period 1998-2003, and 2) firms that contributed soft money to the Democratic
Party during the period 1994-1999. As shown in Table 6, Panel A, we find that Democratic firms
located in areas that would eventually have FNC access in 2000 do not suppress negative
information and subsequently release such information around the 1996 election. These findings
suggest that it is unlikely that unobservable firm characteristics of Democratic firms located in
FNC districts explain our results.
Firm Fixed Effects. To further alleviate potential concerns that FNC access might simply
be capturing unobservable differences, we also run a firm fixed effects model, which controls for
any systematic differences across firms. This specification uses a firm as its own control. As
shown in Table 6, Panel B, we find results consistent with those in Table 4.
– Please insert Table 6 about here –
4.4. Information versus Ideologically Slanted Information
Another potential concern with the results presented thus far is that FNC availability
mainly captures information availability rather than ideological slant in the reporting of this
25
information. Thus, it is possible that our results are driven simply by the availability of an
additional news source. To alleviate this concern, we conduct two additional empirical tests.
Access to Broadband Internet. Around the same time as the diffusion of FNC,
broadband internet became available to more and more U.S. households. Access to broadband
internet is likely to provide access to more, but not necessarily ideologically slanted information.
Thus, we obtain data on internet broadband access from the Center for Policy Informatics20
and
replace our FNC measures with either an Internet indicator, equal to one if the county in which
the firm is located had internet access in the particular year, or Share_Internet, which is the ratio
of households with broadband internet access per county. Using these measures, we rerun Model
1 for the same set of Democratic firms that we use in Table 4. As shown in Table 7, Panel A,
Columns 1 and 2, we do not find that Democratic firms located in areas with broadband internet
access suppress negative information before the election and subsequently release such
information after the election. We also include these additional variables in our base model as
reported in Columns 3 and 4, and continue to observe significant coefficients on our main tests
variables.
Republican Firms and Firms without Political Connections. Second, we rerun Model 1
for the set of firms that contribute soft money to the Republican Party as well as firms that are
not politically connected. Relative to our main sample of Democratic firms, these firms should
be equally affected by the information availability channel but unaffected by the ideological slant
channel. As shown in Table 7, Panel B, we find that both Republican firms as well as firms
without political connections located in areas with FNC availability do not suppress negative
information and subsequently release such information around the 2000 election. These findings
suggest that it is unlikely that access to information explains our results.
20
See https://policyinformatics.asu.edu/broadband-data-portal/dataaccess.
26
– Please insert Table 7 about here –
4.5. Mechanisms
We also examine potential mechanisms through which firms suppress and subsequently
release negative financial information. We focus on real actions that are likely to cause negative
public scrutiny during the election: employment and investment reductions. Prior research shows
that because current employment and investment levels significantly influence voters, politically
connected firms cater to the needs of their politicians by avoiding layoffs and plant closures prior
to an election (Bertrand et al., 2007). Hollister (2016) finds that by the mid 1990’s, public and
media scrutiny over corporate downsizing had reached a level of hysteria, fueled in part by the
1996 election, which brought considerable attention to the issue. We focus on large employment
and investment reductions because the market reaction is significantly more negative for material
reductions (McConnell and Muscarella, 1985; Nixon et al., 2004; Worrell et al., 1991). Our
empirical design follows Kedia and Philippon (2009). These tests allow us to provide evidence
on specific types of information that was temporarily suppressed and, thus, to shed some light on
the potential drivers of changes in stock price behavior we observe.
Workforce Reductions. We conjecture that prior to the 2000 election, Democratic firms
located in areas with FNC availability have particularly strong incentives to avoid releasing news
on workforce reductions given the potential costs that FNC could impose. Consistent with this
prediction, in Table 8, Panel A (columns 1 and 2), we find that Democratic firms with FNC
availability are less likely to have large workforce reductions prior to the election. In the period
after the election, we observe that all Democratic firms (regardless of FNC availability) are more
likely to have large workforce reductions. Democratic firms with FNC availability do not exhibit
stronger evidence of this behavior than Democratic firms without FNC availability. To help rule
out alternative explanations for the pattern of results we document, we also conduct this analysis
27
on Republican firms (columns 3 and 4). Consistent with our negative skewness analysis, we do
not find any evidence that Republican firms located in areas with FNC availability delay
workforce reductions prior to the election.
Investment Reductions. As an alternative operational decision, we also examine the
likelihood of large reductions in property, plant, and equipment for Democratic firms located in
areas with FNC availability before and after the election. As shown in Table 8, Panel B (columns
1 and 2), we find that Democratic firms are less likely to have large reductions in investments
prior to the election if located in an area with FNC availability but are more likely to have such
reductions after the election. This evidence is consistent with the initial suppression and
subsequent release of negative financial information. We do not observe a similar suppress and
release pattern for firms with ties to the Republican Party (columns 3 and 4).
– Please insert Table 8 about here –
4.6. Additional Robustness Tests
Alternative Measure of Ties to Democratic Party. Following Knight (2006), we also use
reports produced by financial analysts (Lehman Brothers, Prudential Securities and International
Strategy and Investment) during the 2000 election campaign as an alternative measure to identify
firms with ties to the Democratic and Republican Party, respectively. These reports identify
firms likely to fare well under Bush (41 firms) and Gore administrations (29 firms) based on the
candidates’ policy platforms. For example, Gore favored price controls and promoted generic
pharmaceuticals (Caremark RX, Express Scripts, TEVA Pharmaceutical, and Watson
Pharmaceutical) while Bush opposed price controls and defended large pharmaceuticals (Bristol
Myers Squibb, Eli Lilly, Merck, Pfizer, Pharmacia, and Schering Plough). Using these reports to
identify firms’ political connections, allows us to link firms to political parties based on policy
28
platforms, ensuring that we capture firms’ economic benefits more directly.21
From this sample
of 70 firms, we exclude foreign firms, firms headquartered in states without FNC data, and firms
without a minimum of five years of financial data. After imposing these constraints, we are left
with 39 total firms (23 Republican firms and 16 Democratic firms).
While the small sample size and resulting lack of power is a major limitation of this
sample, we find results consistent with the results presented in Table 4, as shown in Table 9,
Panel A, Column 1 and 2. In particular, we find evidence that Democratic firms with FNC
availability suppress and release negative financial information during the 2000 election. The
results are stronger in Column 1 as this specification allows for more variation in the FNC
variable (as we use the extent of FNC availability). Column 3 and 4 show the results for firms
with ties to the Republican Party. Consistent with the findings in Table 7, Panel B, we do not
find that Republican firms with FNC availability suppress and release negative financial
information during the 2000 election.
Alternative Sample. Next, we rerun Model 1 using the full sample of firms and include
interaction terms between firms with ties to the Democratic Party, denoted Dem_Firm, the pre-
and post-election indicators, as well as access to FNC. As shown in Table 9, Panel B, our results
are consistent with those reported in Table 4.
Geographic Concentration. As an additional test, we rerun Model 1 using a sample of
firms that are more geographically concentrated in their headquarter state. This analysis helps to
reduce concerns that because firms operate in multiple states and countries, developments in
local media markets are not particularly relevant. We follow Dyreng and Lindsey (2009) and
identify the locations of firms’ subsidiaries using Exhibit 21 of Form 10-K. We define locally
concentrated Democratic firms as those firms with less than 25 subsidiaries outside of their
21
Using this sample of 70 firms Knight (2006) provides evidence that policy platforms are capitalized into prices.
29
headquarter state. As shown in Table 9, Panel C, our results using this reduced sample of
geographically concentrated firms are similar to the results reported in Table 4.
Miscellaneous. As we have data on FNC access as of 1998, 2000, and 2003, in an
additional robustness test (untabulated) we assume the same FNC access in 2003 (our last year)
as in 2000 to hold the firms with FNC access constant. The results are robust. An additional
concern might be that we keep firm-year observations for fiscal year 2000. For these firms, part
of their operations overlapped with the pre- and post-election period. Thus, we rerun Model 1
excluding all observations for fiscal year 2000. The results (untabulated) are robust.
– Please insert Table 9 about here –
5. Conclusions
We document a decrease in negative stock return skewness among Democratic firms
located in areas with access to the Fox News Channel in the year prior to the 2000 election. After
the election, these firms exhibit an increase in negative stock return skewness. This intertemporal
pattern is consistent with the initial suppression and subsequent release of negative financial
information. We interpret this behavior as consistent with the right-leaning FNC increasing the
costs of bad news revelation for left-leaning firms, thereby creating an incentive for these firms
to take actions to avoid negative public and media scrutiny. We do not observe similar patterns
for Democratic firms located in areas without FNC access, firms with ties to the Republican
Party, or non-politically connected firms. We also find that the delay of workforce and
investment reductions are specific actions taken by Democratic firms located in areas with FNC
access to suppress negative financial information.
Our evidence suggests that developments in media markets can influence the behavior of
economic agents in financial markets. We focus on media slant as it represents a wide-spread and
long-standing feature of the information environment in the U.S. Our findings complement prior
30
studies on firms’ response to heightened scrutiny during political events by explicitly shedding
light on the role played by the media in the political process.
31
References
Apollonio, D.E., Raja, R.J.L., 2004. Who gave soft money? The effect of interest group
resources on political contributions. Journal of Politics 66(4): 1134-1154.
Auletta, K., 2001. The Microsoft Verdict, The New Yorker. Available at:
http://www.newyorker.com/magazine/2001/07/09/the-microsoft-verdict.
Ayers, B.C., Ramalingegowda, S., Yeung, P.E., 2011. Hometown advantage: The effects of
monitoring institution location on financial reporting discretion. Journal of Accounting
and Economics 52(1): 41-61.
Baloria, V.P., Klassen, K.J., 2016. Supporting tax policy change through accounting discretion:
Evidence from the 2012 elections. Working paper, Boston College and University of
Waterloo.
Bennet, J., 1998. Clinton says Chinese money did not influence U.S. policy. The New York
Times. Available at: http://www.nytimes.com/1998/05/18/us/clinton-says-chinese-money-
did-not-influence-us-policy.html.
Bertrand, M., Kramarz, F., Schoar, A., Thesmar, D., 2007. Politicians, firms and the political
business cycle: evidence from France. Working paper, University of Chicago.
Bonaparte, Y., Kumar, A., 2013. Political activism, information costs, and stock market
participation. Journal of Financial Economics 107(3): 760-786.
Bushee, B.J., Core, J.E., Guay, W., Hamm, S.J., 2010. The role of the business press as an
information intermediary. Journal of Accounting Research 48(1): 1-19.
Chen, J., Hong, H., Stein, J.C., 2001. Forecasting crashes: Trading volume, past returns, and
conditional skewness in stock prices. Journal of Financial Economics 61(3): 345-381.
Chhaochharia, V., Kumar, A., Niessen-Ruenzi, A., 2012. Local investors and corporate
governance. Journal of Accounting and Economics 54(1): 42-67.
Clinton, J.D., Enamorado, T., 2014. The national news media’s effect on Congress: How Fox
News affected elites in Congress. The Journal of Politics 76(4): 928-943.
Cohen, J., Ding, Y., Lesage, C., Stolowy, H., 2015. Media bias and the persistence of the
expectation gap: An analysis of press articles on corporate fraud. Journal of Business
Ethics, forthcoming.
Cooper, M.J., Gulen, H., Ovtchinnikov, A.V., 2010. Corporate Political Contributions and Stock
Returns. The Journal of Finance 65(2): 687-724.
Core, J.E., Guay, W., Larcker, D.F., 2008. The power of the pen and executive compensation.
Journal of Financial Economics 88(1): 1-25.
32
Dai, L., Parwada, J.T., Zhang, B., 2015. The governance effect of the media's news
dissemination role: Evidence from insider trading. Journal of Accounting Research 53(2):
331-366.
DellaVigna, S., Kaplan, E., 2007. The Fox News Effect: Media Bias and Voting. The Quarterly
Journal of Economics 122(3): 1187-1234.
Drake, M.S., Guest, N.M., Twedt, B.J., 2014. The media and mispricing: The role of the business
press in the pricing of accounting information. The Accounting Review 89(5): 1673-1701.
Dyreng, S.D., Lindsey, B.P., 2009. Using financial accounting data to examine the effect of
foreign operations located in tax havens and other countries on US multinational firms' tax
rates. Journal of Accounting Research 47(5): 1283-1316.
Engelberg, J.E., Parsons, C.A., 2011. The causal impact of media in financial markets. The
Journal of Finance 66(1): 67-97.
Gentzkow, M., Shapiro, J.M., 2010. What drives media slant? Evidence from US daily
newspapers. Econometrica 78(1): 35-71.
Groseclose, T., Milyo, J., 2005. A measure of media bias. The Quarterly Journal of Economics,
120(4): 1191-1237.
Gurun, U.G., Butler, A.W., 2012. Don't believe the hype: Local media slant, local advertising,
and firm value. The Journal of Finance 67(2): 561-598.
Hennes, K.M., Leone, A.J., Miller, B.P., 2008. The importance of distinguishing errors from
irregularities in restatement research: The case of restatements and CEO/CFO turnover.
The Accounting Review 83(2): 1487-1519.
Hollister, M., 2016. Speaking of downsizing: The use of the term “downsizing” in American
news media 1975–2007. Working Paper, McGill University.
Hopkins, D.J., Ladd, J.M., 2014. The consequences of broader media choice: Evidence from the
expansion of Fox News. Quarterly Journal of Political Science 9(1): 115-135.
Hutton, I., Jiang, D., Kumar, A., 2015. Political values, culture, and corporate litigation.
Management Science 61(12): 2905-2925.
Jayachandran, S., 2006. The jeffords effect. Journal of Law and Economics 49(2): 397-425.
Jensen, M.C., 1979. Toward a theory of the press. In Economics and Social Institutions, edited
by K. Brunner. Boston, MA: Martinus Nijhoff: 267–87.
Jin, L., Myers, S.C., 2006. R 2 around the world: New theory and new tests. Journal of Financial
Economics 79(2): 257-292.
33
Kedia, S., Philippon, T., 2009. The economics of fraudulent accounting. Review of Financial
Studies 22(6): 2169-2199.
Kido, N., Petacchi, R., Weber, J., 2012. The influence of elections on the accounting choices of
governmental entities. Journal of Accounting Research 50(2): 443-476.
Knight, B., 2006. Are policy platforms capitalized into equity prices? Evidence from the
Bush/Gore 2000 presidential election. Journal of Public Economics 90(4-5): 751-773.
Kuhnen, C.M., Niessen, A., 2012. Public opinion and executive compensation. Management
Science 58(7): 1249-1272.
McConnell, J.J., Muscarella, C.J., 1985. Corporate capital expenditure decisions and the market
value of the firm. Journal of Financial Economics 14(3): 399-422.
Miller, G.S., 2006. The press as a watchdog for accounting fraud. Journal of Accounting
Research 44(5): 1001-1033.
Miller, G.S., Skinner, D.J., 2015. The evolving disclosure landscape: How changes in
technology, the media, and capital markets are affecting disclosure. Journal of Accounting
Research 53(2): 221-239.
Milyo, J., Primo, D., Groseclose, T., 2000. Corporate PAC campaign contributions in
perspective. Business and Politics 2(1): 75-88.
Mullainathan, S., Shleifer, A., 2005. The market for news. The American Economic Review
95(4): 1031-1053.
Nixon, R.D., Hitt, M.A., Lee, H.U., Jeong, E., 2004. Market reactions to announcements of
corporate downsizing actions and implementation strategies. Strategic Management
Journal 25(11): 1121-1129.
Piotroski, J.D., Wong, T., Zhang, T., 2015. Political incentives to suppress negative information:
evidence from Chinese listed firms. Journal of Accounting Research 53(2): 405-459.
Ramanna, K., Roychowdhury, S., 2010. Elections and discretionary accruals: Evidence from
2004. Journal of Accounting Research 48(2): 445-475.
Reuter, J., Zitzewitz, E., 2005. Do ads influence editors? Advertising and bias in the financial
media. Quarterly Journal of Economics 121(1): 197-227.
Rogers, J.L., Skinner, D.J., Zechman, S.L., 2016. The role of the media in disseminating insider-
trading news. Review of Accounting Studies 21(1): 711-739.
Rutenberg, J., 2000. The right strategy for Fox: Conservative cable channel gains in ratings war,
The New York Times. Available at: http://www.nytimes.com/2000/09/18/business/the-
right-strategy-for-fox-conservative-cable-channel-gains-in-ratings-
war.html?pagewanted=all&_r=0.
34
Shipman, J.E., Swanquist, Q.T., Whited, R.L., 2016. Propensity Score Matching in Accounting
Research. The Accounting Review, forthcoming.
Watts, R.L., Zimmerman, J.L., 1978. Towards a positive theory of the determination of
accounting standards. The Accounting Review 53(1): 112-134.
Worrell, D.L., Davidson, W.N., Sharma, V.M., 1991. Layoff announcements and stockholder
wealth. Academy of Management Journal 34(3): 662-678.
35
Appendix A: Variable Definitions
Variable Description
Primary Dependent Variable
Ncskew
Following Chen et al. (2001) and Jin and Myers (2006), Ncskew is measured
as the third moment of each stock’s weekly residual returns, divided by the
cubed standard deviation of weekly residual returns, times negative one, in
year t.
Mechanism Variables
Workforce_Reduction 1 if a firm’s change in number of employees relative to the average number of
employees employed is in the top quarter of the distribution, zero otherwise.
Investment_Reduction 1 if a firm’s change in gross PP&E relative to the average level of PP&E
employed is in the top quarter of the distribution, zero otherwise.
Political Connections Variables
Soft_Money_Contributor 1 if a firm’s soft money contributions as reported in the Center for Responsive
Politics (CRP) dataset are greater than 0, zero otherwise.
Soft_Money_Dem A firm’s soft money contributions to the Democratic Party as reported in the
CRP dataset.
Share_Dem The percentage of a firm’s soft money contributions allocated to the
Democratic Party.
Election Variables
Election 1 if a firm’s fiscal year end falls in the time period July 31, 1999 to June 30,
2000, zero otherwise.
Post_Election 1 if a firm’s fiscal year end falls in the time period July 31, 2001 to June 30,
2002, zero otherwise.
Fox News Variables
FNC_Subscriptions
The natural logarithm of the number of Fox News Channel subscriptions per
congressional district. We obtained data on Fox News Channel subscriptions
from Joshua Clinton and Ted Enamorado.
FNC_Indicator
1 if Fox News Channel is available in a particular district, zero otherwise. We
obtained data on Fox News Channel availability from Joshua Clinton and Ted
Enamorado.
Firm Visibility Variables
Fortune_500 1 if the firm is covered in the Fortune 500 index as reported in Compustat,
zero otherwise.
Log_Analysts
Natural logarithm of the number of analysts issuing annual earnings forecasts
for firms covered by IBES. Set equal to zero if the firm is not covered by
IBES.
Restate An indicator variable set equal to 1 if the company filed a 10-K restatement in
year t obtained from Hennes et al. (2008), zero otherwise.
Control Variables
Log_Mark_Cap The natural log of market capitalization, calculated as shares outstanding at
fiscal year-end (CSHO) times the share price at fiscal year-end (PRCC_F) as
36
reported in Compustat.
Chg_Sales The percentage change in annual sales (REVT) as reported in Compustat from
year t_1 to year t.
Sigma The standard deviation of weekly excess return in year t.
Turnover The average weekly share turnover in year t.
Lag_Turnover The lagged average weekly share turnover in year t.
Beta A firm’s beta; estimated using weekly returns in year t.
Return A firm’s annual market-adjusted stock return in year t.
Lag_Return A firm’s lagged annual market-adjusted stock return in year t.
Population Population per congressional district as per the U.S. Census Bureau.
Household_Income Household income per congressional district as per the U.S. Census Bureau.
Unemployment_Rate Unemployment rate per congressional district as per the U.S. Census Bureau.
ROA A firm’s return on assets, i.e., Compustat item: IB / Total average assets.
Loss 1 if a firm’s return on assets is smaller than zero, zero otherwise.
Big_4 1 if a firm’s auditor is a Big 4 auditor, zero otherwise.
Mark_to_Book A firm’s market value scaled by firm’s book value, i.e., (Compustat item:
CSHO * Compustat item: PRCC) / Compustat item: CEQ.
37
Figure 1: FNC Availability in 2000
This figure shows FNC availability in 2000. We obtained this figure from Clinton and Enamorado (2014).
Figure 2: Empirical Design around 2000 Election
This figure shows the design of our empirical tests around the 2000 election.
38
Table 1: Sample Selection
Firm-years Firms
Firm(-years) with Compustat and CRSP identifiers 1998-2003 38,103 9,354
Less: Firm(-years) with missing data 5,959 1,298
Less: Firm(-years) of cross-listed firms 3,756 987
Less: Firm(-years) without data on FNC availability 9,739 2,244
Sample of firms 18,649 4,825
Final sample for Model 1 (firms with soft money contributions to
Democratic Party) 670 186
The table displays the sample selection over the period 1998-2003. See Appendix A for variable definitions.
39
Table 2: Descriptive Statistics
Panel A: Pooled samples
All Firms (N=18,649) Democratic Firms Soft Money (N=670)
Variable Mean Std. Median Mean Std. Median
Soft_Money_Contributor 0.036 0.184 0 1 0 1
Soft_Money_Dem 1,857 18,793 0 52,706 85,749 19,000
Share_Dem 0.024 0.141 0 0.675 0.352 0.847
FNC_Indicator 0.472 0.500 0 0.531 0.500 1
Ncskew -0.145 1.335 -0.192 0.059 1.296 -0.118
Log_Mark_Cap 5.202 2.093 5.094 7.494 2.176 7.669
Chg_Sales -0.005 0.909 -0.068 0.070 0.874 -0.052
Sigma 0.046 0.030 0.039 0.031 0.018 0.026
Turnover 0.130 0.146 0.079 0.147 0.139 0.098
Lag_Turnover 0.126 0.141 0.076 0.140 0.139 0.091
Beta 0.736 0.703 0.611 0.825 0.620 0.702
Return 0.139 1.208 -0.070 0.168 0.988 0.024
Lag_Return 0.077 1.119 -0.095 0.187 1.414 0.014
Population 632,479 57,820 620,729 621,883 50,903 619,968
Household_Income 51,954 13,074 50,274 51,655 13,639 49,399
Unemployment_Rate 0.048 0.011 0.049 0.050 0.012 0.052
40
Panel B: Sample of Democratic firms partitioned on Fox News availability
Variable
Fox News
availability (N=356)
Mean (1)
No Fox News
availability (N=314)
Mean (2)
Difference
(1) – (2)
Missing Fox News
availability (N=510)
Mean
Soft_Money_Dem 67,181 35,925 31,256*** 49,274
Ncskew 0.054 0.064 -0.010 0.135
Log_Mark_Cap 7.620 7.379 0.241 7.747
Chg_Sales 0.085 0.087 -0.002 0.008
Sigma 0.030 0.031 -0.001 0.030
Turnover 0.150 0.142 0.008 0.132
Lag_Turnover 0.151 0.127 0.024** 0.126
Beta 0.882 0.772 0.110** 0.843
Return 0.211 0.105 0.106 0.121
Lag_Return 0.229 0.138 0.091 0.061
Fortune_500 0.412 0.411 0.001 0.471
Log_Analyst 2.062 2.010 0.052 2.279
Restate 0.049 0.029 0.020 0.043
Population 614,723 631,093 -16,370*** 678,668
Household_Income 50,667 52,842 -2,175* 45,404
Unemployment_Rate 0.052 0.048 0.004** 0.051
Table 2 shows the summary statistics for all variables used in Model 1 over the period 1998-2003. Panel A shows
statistics for the pooled sample. Panel B shows means separately for Democratic firms with and without FNC access
as well as Democratic firms with missing information on FNC access. Panel B also displays the differences between
the means of the variables. *, **, *** represent significance at the 10, 5, and 1 percent level, respectively. Appendix
A presents variable definitions.
41
Table 3: Pearson (Upper Triangle) and Spearman (Lower Triangle) Correlations in Sample of Democratic Firms
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.
1. Ncskew -0.01 0.01 0.05 -0.07 0.05 0.09 0.03 -0.10 0.01 0.02 0.03 -0.15 0.00 0.06 0.14 0.03
2. FNC_Subscriptions -0.01 0.92 0.21 -0.12 -0.01 0.11 -0.01 -0.03 -0.01 0.05 0.11 0.07 0.06 0.04 0.03 0.04
3. FNC_Indicator -0.00 0.91 0.18 -0.13 -0.01 0.06 -0.00 -0.03 -0.00 0.06 0.09 0.06 0.03 0.00 0.02 0.06
4. Soft_Money_Dem 0.09 0.20 0.17 -0.07 -0.04 0.35 -0.05 -0.21 -0.07 -0.05 0.02 -0.05 -0.01 0.27 0.19 0.03
5. Election -0.16 -0.12 -0.13 -0.04 -0.18 -0.03 0.01 0.08 0.03 -0.03 -0.09 -0.05 -0.03 -0.04 0.02 -0.04
6. Post_Election 0.11 -0.01 -0.01 -0.00 -0.18 0.03 0.03 0.01 -0.03 0.03 -0.03 0.10 0.01 0.01 -0.02 -0.01
7. Log_Mark_Cap 0.17 0.12 0.05 0.44 -0.03 0.04 -0.11 -0.51 0.05 0.06 0.23 0.05 0.08 0.65 0.79 0.05
8. Chg_Sales 0.01 0.00 0.03 0.02 -0.08 0.08 -0.18 0.07 -0.00 -0.02 -0.02 -0.10 -0.04 -0.08 -0.05 0.01
9. Sigma -0.11 -0.08 -0.05 -0.21 0.09 0.03 -0.50 0.14 0.40 0.32 0.27 0.16 0.09 -0.32 -0.33 0.10
10. Turnover 0.08 0.00 0.01 0.10 -0.02 -0.03 0.12 -0.10 0.38 0.76 0.53 0.14 0.33 -0.08 0.17 0.13
11. Lag_Turnover 0.08 0.06 0.08 0.14 -0.02 0.02 0.10 -0.08 0.32 0.84 0.54 0.06 0.25 -0.07 0.21 0.08
12. Beta 0.07 0.12 0.11 0.14 -0.14 -0.08 0.23 -0.06 0.20 0.52 0.53 0.09 0.18 0.02 0.28 0.11
13. Return -0.07 0.06 0.05 -0.04 -0.18 0.13 0.15 -0.23 -0.10 0.03 -0.03 -0.02 -0.02 -0.07 -0.06 0.00
14. Lag_Return 0.12 0.06 0.02 0.00 -0.14 0.12 0.17 -0.16 -0.22 0.06 -0.01 0.09 0.01 -0.01 0.05 -0.00
15. Fortune_500 0.10 0.04 0.01 0.34 -0.03 0.01 0.68 -0.02 -0.34 0.03 0.03 0.05 0.00 0.03 0.57 0.10
16. Log_Analyst 0.17 0.04 0.01 0.34 0.02 -0.04 0.82 -0.14 -0.29 0.27 0.26 0.31 0.01 0.06 0.61 0.04
17. Restate 0.04 0.05 0.07 0.01 -0.04 -0.00 0.04 -0.00 0.09 0.04 0.04 0.08 -0.08 -0.06 0.09 0.05
The table presents the correlation between variables used in the study for the sample of firms contributing to the Democratic Party. The upper triangle shows
Pearson correlations and the lower triangle Spearman correlations. Appendix A presents variable definitions. Bold values indicate significance levels at p<0.1
(two-tailed).
42
Table 4: FNC and News Suppression
(1) (2)
Variables Prediction Ncskew Ncskew
FNC -0.005 -0.059
(0.82) (0.78)
Election 0.110 0.134
(0.59) (0.51)
Election x FNC - -0.037** -0.428**
(0.02) (0.03)
Post_Election 0.078 -0.031
(0.63) (0.86)
Post_Election x FNC + 0.030** 0.521**
(0.04) (0.03)
Log_Soft_Money_Dem 0.039 0.038
(0.53) (0.53)
Log_Mark_Cap 0.025 0.024
(0.66) (0.67)
Chg_Sales 0.059 0.055
(0.45) (0.48)
Sigma -7.042 -7.251
(0.40) (0.39)
Beta 0.026 0.011
(0.85) (0.93)
Turnover 0.116* 0.122*
(0.10) (0.09)
Lag_Turnover -0.037 -0.037
(0.41) (0.38)
Return -0.214*** -0.216***
(0.01) (0.01)
Lag_Return -0.001 -0.000
(0.98) (1.00)
Constant -0.155 -0.172
(0.79) (0.77)
Year FE Yes Yes
State FE Yes Yes
Industry FE Yes Yes
SE clustered by State State
Observations 670 670
R-square 0.174 0.178
43
The table presents results on the relation between FNC availability and news suppression around the 2000 election.
The dependent variable in all models is Ncskew, which is a firm’s third moment of excess daily stock returns scaled
by its cubed standard deviation times minus one. In Column 1, FNC is measured as the natural logarithm of the
number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions. In Column 2,
FNC is an indicator equal to one if Fox News Channel was available in a particular congressional district, and zero
otherwise, denoted FNC_Indicator. All models span the period 1998–2003. The results reported are from an OLS
estimation and use the sample of Democratic firms as defined in Table 1. p-values are displayed in parentheses
below the coefficient estimates and are one-tailed for variables for which we make directional predictions and two-
tailed for all other variables. *, **, *** represent significance at the 10, 5, and 1 percent level, respectively.
Appendix A presents variable definitions.
44
Table 5: Differences in Observable Firm Characteristics
Panel A: Differences in soft money contributions to the Democratic Party – Propensity score
matched sample
(1) (2)
Variables Prediction Ncskew Ncskew
FNC 0.020 0.198
(0.34) (0.41)
Election -0.303 -0.233
(0.31) (0.42)
Election x FNC - -0.037** -0.532**
(0.05) (0.04)
Post_Election -0.070 -0.214
(0.66) (0.29)
Post_Election x FNC + 0.042** 0.655**
(0.01) (0.02)
Chg_Sales 0.059 0.056
(0.50) (0.50)
Sigma -7.185 -7.414
(0.22) (0.19)
Beta -0.068 -0.080
(0.68) (0.63)
Turnover 0.086** 0.087**
(0.03) (0.02)
Lag_Turnover -0.021 -0.016
(0.68) (0.76)
Return -0.393*** -0.390***
(0.00) (0.00)
Lag_Return -0.003 -0.002
(0.96) (0.98)
Constant 0.713*** 0.734***
(0.00) (0.00)
Year FE No No
State FE Yes Yes
Industry FE No No
SE clustered by State State
Observations 366 366
R-square 0.130 0.139
45
The table presents results on the relation between FNC availability and news suppression around the 2000 election.
The dependent variable in all models is Ncskew, which is a firm’s third moment of excess daily stock returns scaled
by its cubed standard deviation times minus one. In Column 1, FNC is measured as the natural logarithm of the
number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions. In Column 2,
FNC is an indicator equal to one if Fox News Channel was available in a particular congressional district, and zero
otherwise, denoted FNC_Indicator. All models span the period 1998–2003. Columns 1 and 2 use a matched sample
of firms with and without FNC availability using propensity score matching. To obtain this sample, we predict in the
first stage (untabulated) the likelihood of being located in a congressional district with FNC availability based on
firm size, soft money contributions to the Democratic Party, two-digit SIC industry code, and year to calculate the
propensity scores using Probit regression estimation with replacement. Second, we match each FNC firm to a non-
FNC firm within the same year, two-digit SIC industry code, size, and soft money contributions to the Democratic
Party using the propensity scores obtained from the Probit regression. To ensure the smallest propensity-score
distance between the treatment and control firms, we apply a caliper matching estimator of 0.15. p-values are
displayed in parentheses below the coefficient estimates and are one-tailed for variables for which we make
directional predictions and two-tailed for all other variables. *, **, *** represent significance at the 10, 5, and 1
percent level, respectively. Variables are winsorized at 1% and 99% levels. See Appendix A for variable definitions.
46
Panel B: Differences in firm visibility
(1) (2) (3) (4)
Variables Prediction Ncskew Ncskew Ncskew Ncskew
FNC -0.002 -0.042 0.023 0.296
(0.92) (0.83) (0.41) (0.25)
Election 0.115 0.146 -0.138 0.062
(0.53) (0.42) (0.79) (0.89)
Election x FNC - -0.035** -0.429** -0.057* -0.963**
(0.02) (0.03) (0.08) (0.01)
Post_Election 0.090 -0.028 0.050 0.011
(0.58) (0.87) (0.84) (0.96)
Post_Election x FNC + 0.032** 0.548** 0.047*** 0.548**
(0.04) (0.02) (0.00) (0.04)
Fortune_500 0.137 0.128
(0.48) (0.51)
Log_Analyst 0.282* 0.294**
(0.06) (0.04)
Restate 0.256 0.270
(0.34) (0.31)
Constant 0.567 0.562 0.640** 0.553**
(0.32) (0.35) (0.03) (0.04)
Controls Yes Yes Yes Yes
Year FE Yes Yes No No
State FE Yes Yes Yes Yes
Industry FE Yes Yes No No
SE clustered by State State State State
Observations 670 670 378 378
R-square 0.186 0.190 0.112 0.118 The table presents results on the relation between FNC availability and news suppression around the 2000 election
controlling for firm visibility. The dependent variable in all models is Ncskew, which is a firm’s third moment of
excess daily stock returns scaled by its cubed standard deviation times minus one. In Column 1 and 3, FNC is
measured as the natural logarithm of the number of Fox News Channel subscriptions per congressional district,
denoted FNC_Subscriptions. In Column 2 and 4, FNC is an indicator equal to one if Fox News Channel was
available in a particular congressional district, and zero otherwise, denoted FNC_Indicator. Columns 3 and 4 use a
matched sample of firms with and without FNC availability using propensity score matching. To obtain this sample,
we predict in the first stage (untabulated) the likelihood of being located in a congressional district with FNC
availability based on firm size, analysts following, Fortune 500 membership, occurrence of a restatement, two-digit
SIC industry code, and year to calculate the propensity scores using Probit regression estimation with replacement.
Second, we match each FNC firm to a non-FNC firm within the same year, two-digit SIC industry code, size,
analysts following, Fortune 500 membership, occurrence of a restatement using the propensity scores obtained from
the Probit regression. To ensure the smallest propensity-score distance between the treatment and control firms, we
apply a caliper matching estimator of 0.15. All models span the period 1998–2003. The results reported are from an
OLS estimation. p-values are displayed in parentheses below the coefficient estimates and are one-tailed for
variables for which we make directional predictions and two-tailed for all other variables. *, **, *** represent
significance at the 10, 5, and 1 percent level, respectively. Appendix A presents variable definitions.
47
Table 6: Differences in Unobservable Firm Characteristics
Panel A: The 1996 Election
(1) (2) (3) (4)
Variables Ncskew Ncskew Ncskew Ncskew
FNCt+4 0.018 0.186* 0.003 0.010
(0.10) (0.05) (0.78) (0.93)
Election_1996 -0.145 -0.140 -0.054 -0.034
(0.23) (0.25) (0.76) (0.85)
Election_1996 x FNCt+4 0.017 0.142 0.018 0.133
(0.19) (0.34) (0.13) (0.40)
Post_Election_1996 -0.075 -0.075 -0.085 -0.080
(0.46) (0.45) (0.63) (0.65)
Post_Election_1996 x FNCt+4 0.013 0.114 0.025 0.250
(0.63) (0.65) (0.50) (0.50)
Constant -1.824*** -1.855*** -1.940** -1. 499**
(0.00) (0.00) (0.03) (0.03)
Controls Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
State FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
SE clustered by State State State State
Observations 999 999 551 551
R-square 0.199 0.199 0.216 0.215
The table presents results on the relation between FNC availability and news suppression around the 1996 election.
The dependent variable in all models is Ncskew, which is a firm’s third moment of excess daily stock returns scaled
by its cubed standard deviation times minus one. In Column 1 and 3, FNC is measured as the natural logarithm of
the number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions. In Column 2
and 4, FNC is an indicator equal to one if Fox News Channel was available in a particular congressional district, and
zero otherwise, denoted FNC_Indicator. Columns 1 and 2 use all firms with soft money contributions to the
Democratic Party. Columns 3 and 4 use the same sample of firms as described in Table 1. All models span the
period 1994–1999. The results reported are from an OLS estimation. p-values are displayed in parentheses below the
coefficient estimates and are two-tailed. *, **, *** represent significance at the 10, 5, and 1 percent level,
respectively. Appendix A presents variable definitions.
48
Panel B: Firm fixed effects
(1) (2)
Variables Predictions Ncskew Ncskew
FNC -0.043* -0.389
(0.10) (0.13)
Election 0.123 0.189
(0.53) (0.34)
Election x FNC - -0.041** -0.533**
(0.03) (0.02)
Post_Election 0.110 0.017
+ (0.45) (0.92)
Post_Election x FNC 0.024** 0.447**
(0.05) (0.04)
Constant -3.131 -3.231
(0.15) (0.16)
Controls Yes Yes
Year FE Yes Yes
State FE Yes Yes
Firm FE Yes Yes
SE clustered by State State
Observations 670 670
R-square 0.094 0.099 The table presents results on the relation between FNC availability and news suppression around the 2000 election.
The dependent variable in all models is Ncskew, which is a firm’s third moment of excess daily stock returns scaled
by its cubed standard deviation times minus one. In Column 1, FNC is measured as the natural logarithm of the
number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions. In Column 2,
FNC is an indicator equal to one if Fox News Channel was available in a particular congressional district, and zero
otherwise, denoted FNC_Indicator. All models span the period 1998–2003. The results reported are from a firm
fixed effects estimation and use the sample of Democratic firms as defined in Table 1. p-values are displayed in
parentheses below the coefficient estimates and are one-tailed for variables for which we make directional
predictions and two-tailed for all other variables. *, **, *** represent significance at the 10, 5, and 1 percent level,
respectively. Appendix A presents variable definitions.
49
Table 7: Information versus Ideologically Slanted Information
Panel A: Broadband Internet access
(1) (2) (3) (4)
Variables Predictions Ncskew Ncskew Ncskew Ncskew
Internet -0.232 -0.175 -0.265 -0.235
(0.42) (0.73) (0.23) (0.23)
Election -0.065 -0.031 0.036 0.064
(0.84) (0.91) (0.90) (0.83)
Election x Internet -0.040 -0.220 0.292 0.267
(0.92) (0.76) (0.29) (0.33)
Post_Election 0.195 0.265 0.162 0.054
(0.40) (0.29) (0.45) (0.82)
Post_Election x Internet -0.093 -0.441 0.239 0.218
(0.71) (0.34) (0.39) (0.44)
Election x FNC - -0.049*** -0.534**
(0.00) (0.01)
Post_Election x FNC + 0.162** 0.496**
(0.04) (0.03)
Constant 0.011 -0.146 -0.061 -0.096
(0.98) (0.76) (0.91) (0.86)
Controls Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
State FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
SE clustered by State State State State
Observations 670 670 670 670
R-square 0.144 0.144 0.178 0.181 The table presents results on the relation between broadband internet as well as FNC availability and news
suppression around the 2000 election. The dependent variable in all models is Ncskew, which is a firm’s third
moment of excess daily stock returns scaled by its cubed standard deviation times minus one. In Column 1 and 3,
Internet is measured as the natural logarithm of the number of broadband internet subscriptions per county, denoted
Internet_Subscriptions. In Column 2 and 4, Internet is an indicator equal to one if broadband internet was available
in a particular county, and zero otherwise, denoted Internet_Indicator. In Column 3, FNC is measured as the natural
logarithm of the number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions.
In Column 4, FNC is an indicator equal to one if Fox News Channel was available in a particular congressional
district, and zero otherwise, denoted FNC_Indicator. All models span the period 1998–2003. The results reported
are from an OLS estimation. p-values are displayed in parentheses below the coefficient estimates and are one-tailed
for variables for which we make directional predictions and two-tailed for all other variables. *, **, *** represent
significance at the 10, 5, and 1 percent level, respectively. Appendix A presents variable definitions.
50
Panel B: Republican firms and firms without political connections
(1) (2) (3) (4)
Variables Ncskew Ncskew Ncskew Ncskew
FNC -0.012 -0.058 0.001 -0.005
(0.38) (0.60) (0.69) (0.86)
Election -0.160 -0.130 -0.174** -0.175***
(0.62) (0.65) (0.02) (0.00)
Election x FNC -0.002 0.001 -0.004 0.009
(0.92) (-0.99) (0.36) (0.82)
Post_Election 0.422*** 0.322*** 0.056 0.061
(0.00) (0.00) (0.36) (0.31)
Post_Election x FNC -0.012 -0.019 0.006 0.074
(0.36) (0.92) (0.20) (0.12)
Constant -0.555*** -0.496*** -1.472*** -0.900***
(0.00) (0.00) (0.00) (0.00)
Controls Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
State FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
SE clustered by State State State State
Observations 1,419 1,419 16,579 16,579
R-square 0.131 0.130 0.093 0.120 The table presents results on the relation between FNC availability and news suppression around the 2000 election.
The dependent variable in all models is Ncskew, which is a firm’s third moment of excess daily stock returns scaled
by its cubed standard deviation times minus one. In Column 1 and 3, FNC is measured as the natural logarithm of
the number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions. In Column 2
and 4, FNC is an indicator equal to one if Fox News Channel was available in a particular congressional district, and
zero otherwise, denoted FNC_Indicator. Columns 1 and 2 use all firms with soft money contributions to the
Republican Party. Columns 3 and 4 use a sample of firms without any political connections measured as no PAC or
soft money contributions. All models span the period 1998–2003. The results reported are from an OLS estimation.
p-values are displayed in parentheses below the coefficient estimates and are two-tailed. *, **, *** represent
significance at the 10, 5, and 1 percent level, respectively. Appendix A presents variable definitions.
51
Table 8: Mechanisms
Panel A: Workforce reductions
(1) (2) (3) (4)
Variables Predictions
Workforce_Red
uction
Workforce_Red
uction
Workforce_Red
uction
Workforce_Red
uction
FNC 0.016 0.182 -0.030 -0.251
(0.48) (0.33) (0.11) (0.19)
Election 0.909 0.926 -0.426 -0.514
(0.50) (0.48) (0.52) (0.46)
Election x FNC - -0.073** -0.796* 0.017 0.359
(0.03) (0.05) (0.68) (0.47)
Post_Election 0.745** 0.630** 1.012*** 0.960***
(0.02) (0.04) (0.00) (0.00)
Post_Election x
FNC +
-0.095** -0.756 -0.060 -0.461
(0.02) (0.15) (0.11) (0.29)
Chg_Sales 0.205 0.207 0.180 0.185*
(0.14) (0.14) (0.10) (0.09)
Log_Mark_Cap -0.044 -0.045 0.004 0.005
(0.54) (0.53) (0.92) (0.90)
Loss 0.879*** 0.869*** 1.386*** 1.383***
(0.00) (0.00) (0.00) (0.00)
Constant -0.232 -0.254 11.614*** 11.610***
(0.82) (0.80) (0.00) (0.00)
Year FE Yes Yes Yes Yes
State FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
SE clustered by State State State State
Observations 620 620 1,415 1,415
Pseudo R-
square
0.125 0.123 0.136 0.135 The table presents results on the relation between FNC availability and workforce reductions around the 2000
election. The dependent variable in all models is Workforce_Reductions, which is equal to one if a firm’s reduction
in employment is in the top quarter, and zero otherwise. In Column 1 and 3, FNC is measured as the natural
logarithm of the number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions.
In Column 2 and 4, FNC is an indicator equal to one if Fox News Channel was available in a particular
congressional district, and zero otherwise, denoted FNC_Indicator. Columns 1 and 2 use all firms with soft money
contributions to the Democratic Party. Columns 3 and 4 use all firms with soft money contributions to the
Republican Party. All models span the period 1998–2003. The results reported are from an OLS estimation. p-values
are displayed in parentheses below the coefficient estimates and are one-tailed for variables for which we make
directional predictions and two-tailed for all other variables. *, **, *** represent significance at the 10, 5, and 1
percent level, respectively. Appendix A presents variable definitions.
52
Panel B: Investment reductions
(1) (2) (3) (4)
Variables Predictions
Investment_Re
duction
Investment_Re
duction
Investment_Re
duction
Investment_Re
duction
FNC -0.059 -0.460 -0.033 -0.260
(0.20) (0.30) (0.32) (0.36)
Election 1.343 1.216 -0.177 -0.251
(0.45) (0.46) (0.74) (0.66)
Election x FNC - -0.179** -1.495** -0.027 -0.098
(0.03) (0.05) (0.68) (0.88)
Post_Election 0.380 0.196 0.583 0.465
(0.58) (0.79) (0.18) (0.30)
Post_Election x
FNC +
0.109** 1.423** 0.004 0.313
(0.05) (0.05) (0.92) (0.53)
Chg_Sales -0.148 -0.143 -0.035 -0.033
(0.44) (0.44) (0.70) (0.73)
Log_Mark_Cap -0.133 -0.131 -0.110 -0.108
(0.24) (0.24) (0.10) (0.11)
Loss 0.678** 0.695** 1.296*** 1.302***
(0.04) (0.03) (0.00) (0.00)
Constant -0.831 -0.923 -2.811*** -2.779***
(0.42) (0.36) (0.00) (0.00)
Year FE Yes Yes Yes Yes
State FE Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes
SE clustered by State State State State
Observations 515 515 1,245 1,245
Pseudo R-square 0.235 0.234 0.153 0.153 The table presents results on the relation between FNC availability and investment reductions around the 2000
election. The dependent variable in all models is Investment_Reduction, which is equal to one if a firm’s reduction
in gross PP&E is in the top quarter, and zero otherwise. In Column 1 and 3, FNC is measured as the natural
logarithm of the number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions.
In Column 2 and 4, FNC is an indicator equal to one if Fox News Channel was available in a particular
congressional district, and zero otherwise, denoted FNC_Indicator. Columns 1 and 2 use all firms with soft money
contributions to the Democratic Party. Columns 3 and 4 use all firms with soft money contributions to the
Republican Party. All models span the period 1998–2003. The results reported are from an OLS estimation. p-values
are displayed in parentheses below the coefficient estimates and are one-tailed for variables for which we make
directional predictions and two-tailed for all other variables. *, **, *** represent significance at the 10, 5, and 1
percent level, respectively. Appendix A presents variable definitions.
53
Table 9: Additional Robustness Tests
Panel A: Alternative measure of ties to Democratic Party
(1) (2) (3) (4)
Variables Predictions Ncskew Ncskew Ncskew Ncskew
FNC -0.017 -0.178 0.033* 0.313
(0.31) (0.32) (0.06) (0.17)
Election -0.380 -0.470 -0.374 -0.278
(0.48) (0.38) (0.62) (0.73)
Election x FNC - -0.062** -0.446 -0.056 -0.795
(0.05) (0.12) (0.27) (0.22)
Post_Election -0.091 -0.126 0.515 0.584
(0.59) (0.42) (0.50) (0.45)
Post_Election x FNC + 0.023** 0.298** -0.014 -0.238
(0.04) (0.05) (0.65) (0.54)
Constant -1.763** -1.525* 0.208 0.351
(0.04) (0.06) (0.87) (0.79)
Controls Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
State FE No No No No
Industry FE No No No No
SE clustered by State State State State
Observations 95 95 130 130
R-square 0.575 0.569 0.284 0.283 The table presents results on the relation between FNC availability and news suppression around the 2000 election.
The dependent variable in all models is Ncskew, which is a firm’s third moment of excess daily stock returns scaled
by its cubed standard deviation times minus one. In Column 1 and 3, FNC is measured as the natural logarithm of
the number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions. In Column 2
and 4, FNC is an indicator equal to one if Fox News Channel was available in a particular congressional district, and
zero otherwise, denoted FNC_Indicator. Columns 1 and 2 use all firms identified as Democratic in Knight (2006).
Columns 3 and 4 use all firms identified as Republican in Knight (2006). All models span the period 1998–2003.
The results reported are from an OLS estimation. p-values are displayed in parentheses below the coefficient
estimates and are one-tailed for variables for which we make directional predictions and two-tailed for all other
variables. *, **, *** represent significance at the 10, 5, and 1 percent level, respectively. Appendix A presents
variable definitions.
54
Panel B: Alternative sample
(1) (2)
Variables Predictions Ncskew Ncskew
FNC -0.005 -0.044
(0.13) (0.17)
Election -0.130** -0.141***
(0.01) (0.00)
Election x FNC -0.004 -0.005
(0.29) (0.90)
Dem_Firm -0.031 -0.058
(0.86) (0.74)
Dem_Firm x FNC 0.001 0.058
(0.95) (0.74)
Dem_Firm x Election 0.030 0.091
(0.90) (0.68)
Dem_Firm x FNC x Election - -0.033** -0.456**
(0.04) (0.03)
Post_Election 0.235*** 0.250***
(0.00) (0.00)
Post_Election x FNC -0.003 -0.058
(0.57) (0.29)
Post_Election x Dem_Firm -0.166 -0.163
(0.42) (0.43)
Post_Election x FNC x Dem_Firm + 0.038** 0.386*
(0.02) (0.08)
Constant -0.782*** -0.778***
(0.00) (0.00)
Controls Yes Yes
Year FE Yes Yes
State FE Yes Yes
Industry FE Yes Yes
SE clustered by State State
Observations 17,604 17,604
R-square 0.099 0.099 The table presents results on the relation between FNC availability and news suppression around the 2000 election
using the full sample of firms as described in Table 1. The dependent variable in all models is Ncskew, which is a
firm’s third moment of excess daily stock returns scaled by its cubed standard deviation times minus one. In Column
1, FNC is measured as the natural logarithm of the number of Fox News Channel subscriptions per congressional
district, denoted FNC_Subscriptions. In Column 2, FNC is an indicator equal to one if Fox News Channel was
available in a particular congressional district, and zero otherwise, denoted FNC_Indicator. All models span the
period 1998–2003. The results reported are from an OLS estimation. p-values are displayed in parentheses below the
coefficient estimates and are one-tailed for variables for which we make directional predictions and two-tailed for all
other variables. *, **, *** represent significance at the 10, 5, and 1 percent level, respectively. Appendix A presents
variable definitions.
55
Panel C: Geographic concentration
(1) (2)
Variables Predictions Ncskew Ncskew
FNC -0.005 -0.299
(0.92) (0.25)
Election -0.135 -0.076
(0.73) (0.86)
Election x FNC - -0.035** -0.277
(0.05) (0.18)
Post_Election 0.039 -0.306
(0.95) (0.31)
Post_Election x FNC + 0.058** 0.756**
(0.05) (0.04)
Constant -2.677* -2.134*
(0.10) (0.07)
Controls Yes Yes
Year FE Yes Yes
State FE Yes Yes
Industry FE Yes Yes
SE clustered by State State
Observations 319 319
R-square 0.276 0.291 The table presents results on the relation between FNC availability and news suppression around the 2000 election.
The dependent variable in all models is Ncskew, which is a firm’s third moment of excess daily stock returns scaled
by its cubed standard deviation times minus one. In Column 1, FNC is measured as the natural logarithm of the
number of Fox News Channel subscriptions per congressional district, denoted FNC_Subscriptions. In Column 2,
FNC is an indicator equal to one if Fox News Channel was available in a particular congressional district, and zero
otherwise, denoted FNC_Indicator. Columns 1 and 2 use a sample of firms with geographically concentrated
operations. We follow Dyreng et al. (2009) and identify the locations of a firms’ subsidiaries using Exhibit 21 of
Form 10-K. We define geographically concentrated Democratic firms as those firms with less than 25 subsidiaries
outside of their headquarter state. For the average Democratic firm these 25 subsidiaries are located in the
headquarter state and two states outside of the headquarter state. All models span the period 1998–2003. The results
reported are from an OLS estimation. p-values are displayed in parentheses below the coefficient estimates and are
one-tailed for variables for which we make directional predictions and two-tailed for all other variables. *, **, ***
represent significance at the 10, 5, and 1 percent level, respectively. Appendix A presents variable definitions.